Nature | Vol 586 | 15 October 2020 | 409showed complete rescue of LTM (Fig. 2f). Next, we tested 4Ekd animals
in the differential threat-conditioning paradigm (Fig. 1a). SOM.4Ekd
mice learned equivalently to SOM.GFP mice (Extended Data Fig. 6d–f ).
During LTM, SOM.4Ekd mice displayed a selective impairment in the
conditioned-threat response to the CS+ but showed a normal safety
response to CS− and a normal cue discrimination index (Fig. 2g–i).
PKCδ.4Ekd mice also acquired differential threat associative memory
normally (Extended Data Fig. 6g–i). However, PKCδ.4Ekd mice showed a
selective impairment in the conditioned-safety response to CS−, despite
a normal conditioned-threat response to CS+ (Fig. 2g, j), which led to a
sub-optimal cue discrimination index for PKCδ.4Ekd animals (Fig. 2k).
Both SOM.4Ekd and PKCδ.4Ekd animals displayed negligible baseline
freezing during pre-CS (Extended Data Fig. 6j–m). Overall, these results
show that prevention of cap-dependent translation in SOM and PKCδ
INs results in selective impairments in conditioned threat and safety
responses, respectively.
To understand the contribution of time-limited de novo protein
synthesis during the initial consolidation window following learning,
we applied a knock-in mouse-based chemogenetic strategy^10 to express
Cre-dependent and drug-inducible double-stranded RNA-activated
protein kinase (iPKR) in SOM and PKCδ INs (Fig. 3a). Because the iPKR
mouse line also enables Cre-dependent expression of eGFP-tagged
ribosomal subunit L10, we detected soma-localized GFP in the CeL
SOM and PKCδ neurons of SOM.iPKR and PKCδ.iPKR mice, respectively
(Fig. 3b). In vivo infusion of Asunaprevir (ASV), the drug inducer of iPKR,
substantially increased the phosphorylation of eIF2α (S51) in SOM and
PKCδ INs in the CeL (Extended Data Fig. 7a, b). We then exposed the
animals to the differential threat-conditioning paradigm as described
above, but restricted cell-type-specific inhibition of protein synthesis
to the initial consolidation period by infusing ASV into the CeL imme-
diately after training (Fig. 3c). Although all mice learned equivalently
during training (Extended Data Fig. 7c–h), we found remarkably diver-
gent memory deficits in SOM.iPKR and PKCδ.iPKR mice. Similar to
the 4Ekd approach, we found that blocking general translation with
increased eIF2α phosphorylation in SOM INs impaired the freezing
response to CS+ while keeping the safety response and cue discrimi-
nation intact (Fig. 3d–f). On the other hand, blocking general transla-
tion with increased eIF2α phosphorylation in PKCδ INs resulted in ana b SOM.4Ekd PKCδ.4EkdGFPWT PKCδ.4Ekd02040608010001020304001020304050SOM.4Ekd
*** **** ***WT SOM.4Ekd020406080(^100) *
LT M1LTM2 LTM1LTM2
0
20
40
60
80
100
ef
Freezing (%)
SOM.4Ekd
PKCδ.4Ekd
WT
CS+ CS–
0
20
40
60
80
100 NS
–0.2
0.0
0.2
0.4
0.6
0.8
1.0
PKCδ: GFP 4Ekd
g jk
PKC
δ.GFP
PKC
SOM: GFP 4Ekd δ.4Ekd
0
20
40
60
80
100 **
h CS+
CS–
i
0
0
0
40
40
40
(^220) Time (s) 370
Discrimination index
NS
Motion index (AU)
CS+
CS–
SOM.GFPSOM.4Ekd
0.0
–0.2
0.2
0.4
0.6
0.8
1.0
d
c
Freezing (%) Freezing (%)
Freezing (%) Freezing (%)
CAG pr tTA
Promoter cre
TREGFP
Col1a1 shmir-eIF4E
CeL
Sst or Prkcd
Off dox
AAV
changeDiet
Hab.TrainingLT MRe-trainLTM2
–21 d –14 d–24 h
On dox Off dox^0 +24 hOn dox+14 d
+15d
Pre-CS
CS1CS2
Pre-CS
CS1CS2
Pre-CS
CS1CS2 WT
SOM.4EkdPKC
δ.4Ekd
Discrimination index
Fig. 2 | Cell type-specif ic inhibition of cap-dependent translation in CeL
INs. a, Intersectional chemogenetic strategy for knocking down eIF4E in CeL
INs. pr, promoter. b, Brain-region- and cell-type-specific expression of GFP.
shmir-eIF4E in CeL SOM and PKCδ INs. c, Behaviour paradigm for simple and
differential cued threat-conditioning. Hab., habituation. d, Normal memory
acquisition for simple threat conditioning in wild-type (WT), SOM.4Ekd and
PKCδ.4Ekd groups. WT, F(2,33) = 10.44, P = 0.0003; SOM.4Ekd, F(2,30) = 16.26,
P < 0.0001; PKCδ.4Ekd, F(2,21) = 13.46, P = 0.0002. e, SOM.4Ekd mice display
significantly impaired LTM compared to wild-type or PKCδ.4Ekd mice.
F(2,28) = 6.41, P = 0.0051. d, e, n = 12 (WT), 11 (SOM.4Ekd) and 8 (PKCδ.4Ekd)
mice. f, Re-training SOM.4Ekd mice after placing on doxycycline
(dox)-containing diet for 14 days rescued the memory deficit. Effect of drug,
F(1,13) = 12.33, P = 0.0038; effect of genotype, F(1,13) = 21.13, P = 0.0005. n = 7
(WT) and 8 (SOM.4Ekd) mice. g, Representative motion traces during
differential threat LTM test. h, SOM.4Ekd mice are impaired in CS+ threat LTM
compared to SOM.GFP controls, but show equivalent safety response to CS−.
Effect of genotype, F(1,44) = 6.68, P = 0.013; effect of CS, F(1,44) = 58.41,
P < 0.0001. n = 12 mice per group. i, Normal cue discrimination index for
SOM.4Ekd mice compared to controls. P = 0. 37 7. j, PKCδ.4Ekd mice are
impaired in safety LTM to CS− despite showing normal threat LTM to CS+. Effect
of genotype, F(1,34) = 4.17, P = 0.049; effect of CS, F(1,34) = 28.60, P < 0.0001.
n = 9 (PKCδ.GFP) and 10 (PKCδ.4Ekd) mice. k, Discrimination index for cued
threat is impaired in PKCδ.4Ekd mice. P < 0.0001. n = 9 (PKCδ.GFP) and 10
(PKCδ.4Ekd) mice. d, f, Repeated measures one-way ANOVA with Bonferroni’s
post-hoc test; e, one-way ANOVA with Bonferroni’s post-hoc test; h, j, two-way
ANOVA with Bonferroni’s post-hoc test; i, k, unpaired t-test, two-tailed.
Mean ± s.e.m. *P < 0.05, *P < 0.01, P < 0.001, ****P < 0.0001. NS,
nonsignificant. Scale bar, 50 μm.